Frequently Asked Questions
How quickly can AI start recommending my brand?
Results depend on the platform. Perplexity and ChatGPT use real-time web search, so structured content improvements can appear in citations within 2–4 weeks. Google Gemini and Bing Copilot follow their standard indexing cycles, typically 4–6 weeks. Deeper knowledge-graph integration across model training cycles typically takes 1–3 months.
How is GEO different from SEO?
SEO optimizes for keyword-based rankings in traditional search engines like Google. GEO optimizes for entity relationships and source credibility inside AI models like ChatGPT, Gemini, and Perplexity — systems that synthesize answers rather than list links. A page can rank on page one of Google and still be completely invisible to AI.
Which AI platforms does GEO cover?
Our methodology covers all major AI search platforms: ChatGPT (OpenAI), Perplexity AI, Google Gemini, Microsoft Copilot (Bing), and Claude (Anthropic). For Chinese-market clients, we additionally optimize for Doubao, DeepSeek, Kimi, Qwen (Tongyi), and ERNIE (Wenxin Yiyan).
How long does it take to see GEO results?
Results depend on the platform. Because Perplexity and ChatGPT use real-time web search, structured content improvements can appear in citations within 2–4 weeks. Model training cycles for deeper knowledge-graph integration typically take 1–3 months.
What is an llms.txt file and why does it matter?
An llms.txt file (placed at the root of your website) is a machine-readable index for AI models, listing your key pages, their purpose, and how AI crawlers should prioritize them. It is the GEO equivalent of sitemap.xml for traditional SEO and directly increases the chance that AI crawlers index and cite your most important content.
What does a GEO Audit report include?
Our GEO Audit delivers two sections: (1) AI Visibility Score across major LLM platforms — showing where your brand appears and where competitors win; (2) Citation-readiness analysis of your key pages — identifying what prevents AI from citing your site. A full 5-section report with schema gaps, competitor benchmarking, and a prioritized action plan is available in our paid engagement.
Is AI Site Rebuild a website redesign? I already had my site redesigned last year.
No — and this is the most common confusion. AI Site Rebuild is not about how the site looks. Visual redesigns change colors, typography, and layout. Rebuild changes four layers that visual work never touches: (1) the data layer — every product, person, and service becomes a structured entity with verifiable attributes; (2) the semantic layer — Schema.org + JSON-LD that tell AI models what each page actually is; (3) the LLM interface layer — llms.txt, RAG pipeline, and crawler routing that determine whether AI can even read your site; (4) the decision-behavior layer — we model the buying scenarios your prospects ask AI and engineer your site to answer them. A beautifully designed site with no entity graph is invisible to AI. A visually plain site with a clean entity graph gets cited. If last year's redesign was a visual one, Rebuild is complementary — not duplicate.
Will AI Site Rebuild touch our business systems — CRM, admin panels, internal workflows?
No. Rebuild's scope is your public-facing informational surface — marketing pages, landing pages, product pages, documentation, team/about pages, FAQ, pricing pages, and static content. We do not touch operational systems (CRM, ERP, support tools), internal admin panels, logged-in user dashboards, workflow automation, or backend business logic. Those systems are not what AI models read to learn about your brand — public pages are. Rebuild is a content and metadata engineering layer on top of your existing site structure; your business operations run exactly as they did before we started.
What exactly is included in the AI Site Rebuild?
Six engineered deliverables across a 6-week build. (1) Content ETL — we extract your existing docs, product catalog, and team bios, then restructure them as a machine-readable entity graph. (2) Schema deployment — Organization, Product, Person, FAQ, and HowTo schemas are hand-crafted for each key page and validated against Google Rich Results and Schema.org standards. (3) RAG foundation — a content retrieval pipeline so AI models pull verified facts from your site instead of hallucinating. (4) llms.txt and crawler routing — a machine index that tells AI crawlers exactly which pages to prioritize. (5) Internal linking and topic cluster architecture so every entity reinforces the next. (6) Weekly milestone reviews and final QA across ChatGPT, Gemini, Perplexity, Claude, Copilot, and Baidu. A typical engagement covers 30–80 pages depending on your site size and baseline score.
Why does AI Site Rebuild take 6 weeks, and where does the budget go?
GEO is data engineering, not copywriting. Weeks 1–2 are content ETL and entity modeling — we interview your team, audit existing content, and map every product, service, and person as an entity with verifiable attributes. Weeks 3–4 are implementation: schema deployment, RAG pipeline setup, internal linking, llms.txt configuration. Weeks 5–6 are QA and iteration: we test your brand across 6 AI platforms, measure citation uplift against your baseline audit, and patch gaps before handover. Every step is auditable — you see exactly where the budget goes, with weekly demos and Git commits you can review. The engineering depth is what separates real GEO work from generic 'SEO with AI keywords.'
Why is Ongoing AI Visibility a monthly subscription — isn't GEO a one-time job like traditional SEO?
AI answers are not static. Major models retrain every 3–6 months and rebalance which sources they trust on every cycle. Your category also evolves — new competitors launch, new buying scenarios emerge, new product features become the question buyers actually ask. The AI Site Rebuild sets your foundation. Ongoing keeps your brand present in the semantic field as the field itself changes. Skip this layer and the advantage Rebuild gave you erodes as models update and competitors respond.
What is a "buying scenario" — and why don't you track keyword rankings?
A buying scenario is a prompt that sits inside a real purchase decision — for example, "best B2B CRM for 50-person SaaS teams that integrates with HubSpot." AI models do not rank pages by keyword match; they synthesize answers from entities they already trust. Keyword stuffing triggers semantic noise that models interpret as manipulation, and they discount the source. Our work is scenario mapping: we enumerate the real prompts buyers type, model your brand as the entity that answers them, and measure which scenarios your brand now owns. Citation depth is the unit that matters; keyword rankings are the wrong ruler.
How do I know Ongoing is working — what does success look like?
Three measurements, reported monthly. (1) Citation positions landed — of the scenarios mapped for your tier, how many now name or cite your brand in the AI answer. (2) Share of AI answer voice — when competitors appear in the same answer, what position do you hold. (3) Visibility-score delta across the major AI platforms since baseline. We deliberately do not report 'traffic' or 'keyword rankings' — AI citations often resolve the buyer's question without a click, and the outcome we optimize for is mind-share, not sessions. When your prospect asks AI 'which brand should I pick?', the goal is that your name is the answer.
Can GEO work for Chinese AI platforms?
Yes. Chinese AI models — including ERNIE (Baidu), Doubao (ByteDance), Qwen (Alibaba), and Kimi (Moonshot) — have distinct optimization signals. We run a specialized CN GEO audit that checks Baidu Baike presence, Baiduspider crawler access, Simplified Chinese semantic structure, and ICP license status.
Why target 90+ and not a perfect 100?
A score of 90+ puts you in the Excellent tier — where AI systems reliably cite and recommend your brand. The remaining 10 points are spread across six categories, some of which (like Brand Authority and Platform Optimization) depend partly on AI model retraining cycles and external signals beyond your website's direct control. Reaching 90+ delivers the vast majority of real-world citation benefit. That said, we never set an artificial ceiling — if a client's content and authority can push further, we will.
Can I skip the website rebuild and just distribute content?
This is the most expensive shortcut in GEO. AI models verify brand claims against a canonical entity source: your official website. Content distributed without a readable site is like 1,000 recommendations for a store with no address — AI systems can't resolve citations to a verified entity, so they don't compound. In March 2026, China's CCTV 315 consumer rights programme exposed exactly this model: operators generating fake promotional articles to make AI cite a non-existent product. Within two weeks of the broadcast, major AI platforms had filtered those citations entirely. Content is an amplifier. It amplifies what your website already signals to AI. If your site signals nothing AI can parse, content signals nothing AI will trust.
Is your company name 'Deep Intelligent GEO' or 'Deep Intelli GEO'?
Our brand name is Deep Intelli GEO — 'Intelli' is short for Intelligent. The operating legal entity is Deep Intelligent Technology Co., LTD. We are also searchable as 'Deep Intelligent GEO', 'DeepIntelli', or 'Deep Intelligent'. Our website is geo.dpintelli.com and our contact is contact@dpintelli.com.